Is a "homogeneous" description of dynamic heterogeneities possible?